Least mean squares channel estimation for downlink Non-Orthogonal Multiple Access

Abstract

Wireless communications as a field of interest has evolved immensely with a seemingly ever-growing impact on modern life. Increasing service demands and innovations posed a need for the fifth generation (5G) wireless mobile network. Non-Orthogonal Multiple Access (NOMA) has been considered as an essential principle to realize some of the 5Grequirements. NOMA is a multiple access scheme for accommodation of more users than the available resources. For reliable communication in wireless mobile systems, there is a need for channel estimation(CE). This dissertation considers CE for NOMA-based wireless communication systems. A simplified, two-user single-output-single-input (SISO)-NOMA scenario is considered in order to focus attention on the development of efficient CE approaches. These CE schemes can then be extended to a NOMA model that accommodates a large number of users and multiple antennas. Four pilot aided CE approaches are presented in this dissertation. The first two approaches consider the constant step-size least mean squares (CSS-LMS) algorithm, the self-adaptive variable step-size LMS (VSS-LMS) algorithm and the self-adaptive VSS-LMS with adaptive speed (VSS-LMS-AS) algorithm. The last two approaches employ the CSS-LMS algorithm. The first approach entails a scenario whereby all the pilots are known to both users. In using the CSS-LMS algorithm, closed form expressions are derived for the optimal step-size and the corresponding error variance. The rest of the approaches assume partial knowledge of the training information. The second approach is the averaging sum-based CE scheme, while the third approach is a case-wise LMS-based CE and detection technique. A successive interference cancellation (SIC)-inspired iterative CE and detection approach is proposed as the fourth scheme. For all four schemes, the simulation results were in accordance with the theoretical analysis. For both the all-known-pilots scheme and the averaging sum scheme, the CSS-LMS algorithm is the best algorithm in terms of CE speed. This is in comparison to the VSS-LMS and the VSS-LMS-AS. Power allocation was found to affect the CE performance in terms of both CE speed and accuracy for when using the averaging sum-based CE scheme. For the last two approaches detection error was shown to negatively affect the CE performance. As a result, motivating for detection schemes with a sufficiently low probability of error